Compute Kendall's Tau of an Ali-Mikhail-Haq ("AMH") Archimedean copula
with parameter theta. While that's analytically given
explicitly, as
$$1-\frac{2((1-\theta)^2\log(1-\theta) + \theta)}{3\theta^2},$$
for th$=\theta$;
numerically, care has to be taken when $\theta \to 0$, avoiding accuracy
loss already, for example, for $\theta$ as large as theta = 0.001.
Usage
tauAMH(th)
Arguments
th
numeric vector with values in $[0,1]$.
Value
a vector of the same length as th ($= \theta$), with
$\tau$ values in $[0, 1/3]$, of
$\tau_A(\theta) = 1 - 2(\theta+(1-\theta)^2\log(1-\theta))/(3\theta^2)$, numerically
accurately, to at least around 12 decimal digits.
Details
For small th ($=\theta$), we use Taylor series
approximations of up to order 7,
$$\tau_A(\theta) = \frac{2}{9}\theta(1 + \theta(\frac 1 4 +
\frac{\theta}{10}(1 + \theta(\frac 1 2 + \theta \frac 2 7)))) + O(\theta^6),$$
where found that dropping the last two terms (e.g., only using 5 terms
from the $k=7$ term Taylor polynomial) is actually numerically advantageous.